Planning | Description | Requirements | UI Mockup | Triples Examples | Queries Examples

Table of Contents

Planned Stack

• Infrastructure – SHARED — code that can be shared by all universities
  •  RDF Models for Triples:  RDF Model classes and RDF Vocabularies (beyond those defined in ruby-rdf/rdf/lib/rdf/vocab) are defined in a local Ruby on Rails app.  
    •  Developing with Ruby:  2.1.2
    •  Developing with Rails:  4.1.1
    •  Gems:  Active-Triples
      •  The RDF Model classes extend the Resource class defined in the Active-Triples gem.
      •  NOTE: Active-Triples brings in code from several other projects.  I’m not sure of the full list, but ruby-rdf/rdf is certainly used by Active-Triples.
  •  SOLR Index
    •  Triples will be set up to go to a SOLR index using code currently in or soon to be developed code in Active-Triples gem.  I know there is some SOLR support, but I don’t know how extensive it is.  For example, we will want to put some information in our SOLR index that isn’t going into the triplestore and may need to expand the code to accomplish this.
    •  UI will query the SOLR index to retrieve data to display on the screen
  •  Triplestore
    •  I am currently developing over a simple triplestore in sqlite3.  We will be moving the final version to a fuller featured triplestore long term.  The triplestore is configured as part of the ActiveTriple gem and should not be difficult to move to a different triplestore system.  At this point, it is envisioned that this will be a standalone triple store that is separate from the main library catalog triplestore that will be created through the conversion process.   
•  Infrastructure – Potentially Shared – code that may be able to be shared by all universities
  •  Conversion of university id (e.g. Cornell netid) into a VIVO URI.  Each university may need to write this utility for their IDs with a conversion to the URI of choice for that university.
•  User Interface — NOT SHARED — all that follows is Cornell specific
  •  Authentication Code
    •  For Cornell, we will connect into our Shibboleth login system for authentication.
  •  UI based on mockups in the wiki will be integrating the infrastructure code into the Cornell catalog system which is implemented over backlight.

Development Status

Ok, best laid plans of mice and men.  Implementation is not following the Sprints - Planned Development Approach.  Here are where things are currently...

• Infrastructure – SHARED
  
  • RDF models and vocabularies are 90% complete, but need a little more testing and perhaps tweaking.
    • 90% - models and vocabularies are complete and base ad-hoc testing in rails console
    • 80% - models are documented in Triples Examples wiki page to facilitate conversations with ontology group
      • outstanding issues
        • owner, contributors - Need to look into PROV ontology (per suggestion from Paolo)
        • lists - There are still discussions about whether to represent lists as ORE or Collections.  I plan to wait until I am further along and know more about the implications for implementation for each model.
    • 25% - currently writing rspec tests for models.  The test for the first model is close to being complete, but not sure what issues will arise when it runs.  The others should be quick to write since they are going to be very similar to the tests for the first model.
      
  • SOLR index - not begun beyond glancing at the code in ActiveTriple that works with SOLR
    • 5% - need to explore ActiveTriple code specific to SOLR in more detail and play around to see if it can already do what we need
    • 80% - queris are documentd in Queries Examples wiki page to facilitate conversations with ontology group
      • Initially, these were defined as SPARQL queries, so they need to be rewritten as SOLR queries.
        
  •  Triplestore - currently using sqlite3
    • 5% - need to explore fuller featured triplestore (DEPENDENCY:  Prefer to wait and see what Rebecca decides to use for her work.)
• Infrastructure - Potentially Shared
  • 10% - I have information from Jim Blake about how to get the information from VIVO, but there isn't a utility supported by VIVO that provides the URI. 
    
  • 0% - Write a utility that converts a Cornell netid into a VIVO URI.
• User Interface – NOT SHARED
  • Authentication Code
    • 5% - I've spoken with folks about how Cornell does authentication and looked over code that is similar to what will need to be done.
  • UI
    
    • 20% - I have cornell-blacklight up and running and have identified several places where we will be inserting code.
      
      
• ActiveTriples development
  • I have been corresponding with Tom (the developer of ActiveTriples) to understand his vision, current needs for development.  Once I grok where things are, I can start to see if there are shortcomings that limit our goal of ActiveTriples being easy to use in a Hydra stack without Fedora.
  • On the practical side, I have been creating a reference guide for ActiveTriples as I learn how the code works.  There is very minimal doc at this point and I am having to read class and rspec code to figure out how things work.  A reference guide will be an extremely valuable tool for future developers.
    

Sprints - Planned Development Approach

Sprint 1:  Add a single work from an individual work page to a single supported virtual collection named My Virtual Collection.

Sprint 2: Use ActiveTriples instead ActiveFedora

TBA

Ontology Questions

NOTE:  Many of these questions have been answered.  I haven't had time to update the answers here.  Most of the Ontology Questions are on the Triples Examples page including answers and current state of the discussion.

User

• How to represent a user?
  • Possibility <foaf:person>
• How to represent ownship/authorship of a user?
  • <co:list> <dc:creator> <foaf:person>
• How to represent user role to list type relationships
  • <foaf:person> owns <co:list>
  • <foaf:person> canEdit <co:list>  –  facilitates collaboration
  • <foaf:person> canView <co:list>
• How to represent user to group to list type relationships
  • <foaf:person> isIn <group>
  • <group> canEdit <co:list>
  • <group> canView <co:list>
• Are groups and roles in the triplestore?  
  • Or are they maintained separately and in the triple store on as a uri to the group?  
  • Issue:  Any privacy issues with user data being repeated in our triplestore?
  • Issue:  What information do we have access to and permission to put in a public triplestore?

Virtual Collection

• How to represent metadata about the virtual collection?
  • basic metadata
    • title – use <rdf:label>?
    • description
  • controlled vocabulary metadata
    • list of selected vocabularies (ordered list) – use <co:List>?
    • what metadata about each controlled vocabulary will be stored in the triplestore?
      • name – use <rdf:label>?
      • description
      • id:label for each term in the vocabulary – use <co:List>?
    • Are these stored in the main triplestore or as a test file in the file system that is accessible via URL? 
      • NOTE: Questioning Authority has them as files in the rails app structure at /config/authorities, which is not accessible via URL and has a non-triple format.
  • access controls - How to record metadata about?  How to limit access to triples related to virtual collections based on access level?
    • private (only the owner & collaborators can view)
    • shared (semi-private, only those with the access URL can view)
    • public (everyone can view)
    • collaborative (multi-editors)

UI Questions

NOTE:  Some of these questions have been answered.  I haven't had time to update these, so the questions may be stale.  Hopefully, I will update soon.

• How does user stayed logged in to facilitate multiple searches to build a single Virtual Collection? 
  
  • Currently, every time a user clicks My Account, a prompt for logging in is displayed, which implies users are logged out as soon as they leave the My Account page.  (Ask Adam the history of this.)
  • Can we add the notion of a public and private computer?
  • Can we keep an IP address (or other identifier) for a computer that allows us to know this user is accessing the system from a previously authorized private computer?
  • What default length of time forces a log out?  Is this settable by the user?
• How to find a Virtual Collection to watch?
• How to search only within the user's owned Virtual Collections?
• How to search only within the user's watched Virtual Collections?
• How are user public Virtual Collections exposed to the public?
• How are works from external library catalogs searchable and available for inclusion in a Virtual Collection?  (Potential Demonstration B)
• The UI Mockup implies that items in a Virtual Collection can be dynamically sorted (show in title A-Z).  Are dynamic sorts sufficient (title A-Z/Z-A, author A-Z/Z-A, year ASC/DES)?  Potential Demonstration C states that the user can apply an order to items.

Status of Planning Process

• UI Mockup is fairly well developed for Potential Demonstrations A and D.  See UI Questions above for areas that may not be represented in the mockup.
• Requirements are fairly well developed for Potential Demonstrations A, C, and D.
• See implementation status above in section Development Status.